Shanghai AI Lab open-sources Sage-S2.1Flash, tops MLE-bench at 90.3
机器之心 · wechat · 2026-11-16
Shanghai AI Laboratory, with Fudan, PKU and other institutions, released and open-sourced Sage-S2.1Flash, a scientific multimodal LLM aimed at AI-driven research (AI4AI).
- Scores 90.3 on MLE-benchLite, 72.2 on OfficeQAPro, 25.0 on HarveyLAB, and 37.1 on SpreadsheetBench2 — beating DeepSeek-V4.1-Flash-1017 and Claude Opus 5 in those comparisons; 4 firsts and 7 seconds across 33 agentic benchmarks
- Demonstrates real research-software workflows: 3D bone reconstruction in 3DSlicer, editable building scenes in Blender, Poyang Lake water analysis in QGIS, parametric engine assembly in CAD — all outputs remain editable and verifiable
- Accepts 10+ scientific input types: molecular graphs, crystal structures, protein/DNA/RNA sequences, spectra, EEG and more
- Priced at as low as 1% of the comparable GLM-5.3-Flash during the launch promotion
Available on ModelScope and Hugging Face with an open tech report.
More from Models
- Every's vibe check: Gemini 4 Carbon codes better but judgment lags Fable 5.2 — every · 2026-11-16
- Fable 5.2 vs Astra 6 on real PCB layout: 2 clean iterations vs 30 failed ones — ilovemydickuwu · 2026-11-16
- Google's rumored Gemini 4 Oxygen Flash model turns heads at design work — feblet · 2026-11-16
- Shanghai AI Lab's Sage S2.1 Flash tops 4 of 15 scientific benchmarks — pmttyji · 2026-11-16
- DeepSeek free-tier limits spark one-star review revolt as users treat free AI as entitlement — teortaxesTex · 2026-11-16
- User claims Fable 5.2 opens obvious security holes, suspects it was deliberately nerfed — wavefnx · 2026-11-16